On the morning of October 10, 2024 vivo In the opening speech of the developer conference, Vivo officially released its new self-developedBlue Heart Modelmatrix, comprehensively upgrading the language model and end-side model capabilities, and bringing vivo's self-developedVoice big model, image macromodeling, and multimodal macromodeling.
It is understood that the new Blue Heart Big Model Matrix includes Language Big Model, End Side Big Model, Speech Big Model, Image Big Model, and Multimodal Big Model. vivo has launched the new 3 billion Blue Heart End Side Big Model 3B.Officially, the Blueheart 3B can surpass the industry's 7B-9B models in terms of dialogue writing, summary summarization, and information extraction.Compared with BlueCenter 7B, BlueCenter 3B has an extreme performance increase of 300%, power consumption optimization in balanced mode of 46%, and memory reduction of 63%. Compared with Blueheart 7B, Blueheart 3B has an extreme performance increase of 300%, a power consumption optimization of 46% in balanced mode, a memory reduction of 63%, and an extreme word output speed of 80 words/s, with a system power consumption of only 450mA, and a memory occupation of only 1.4GB. In both SuperCLUE and Equal Eval's evaluation of the comprehensive capability of small models, Blueheart 3B ranked No. 1. BlueCenter 3B ranked No. 1 in SuperCLUE and Equal Eval's Small Model Comprehensive Capability Evaluation, and was also certified by Tyrell Labs for its end-side Large Model Comprehensive Capability and AIIA Security Prevention Capability.
vivo also unveiled a new self-developed Blue Heart voice grand modelIt can accurately understand natural semantics, simulate the human voice, express emotions, and support simultaneous interpretation.
vivo has also upgraded the Blue Heart Image & Multi-Modal Grand ModelingThe blue heart image large model, which strengthens the fusion of Chinese characteristics and oriental aesthetics generation ability, is officially said to be one of the most understanding of the Chinese context of the image model in China, differentiated from other large models, support for the national style of ink and support in the process of generating pictures to increase the creation of Chinese characters related to the; blue heart multimodal large model, this year, upgraded the ability to perceive and understand the vision, able to understand the cell phone screen in greater depth, and more smooth and natural in the The real-time dialog of video streaming is also more smooth and natural. BlueCenter's Speech Big Model, Image Big Model and Multimodal Big Model have ranked first in many professional lists.